├── Zurich_Bicycle_Dataset/
│ ├── DSCN2571/
│ ├── GOPR0200/
│ ├── GOPR0255/
│ ├── GOPR0265/
│ ├── GOPR0366/
│ ├── GOPR0369/
│ ├── GOPR0382/
│ ├── GOPR0386/
│ ├── LICENSE.mit.md
│ ├── README.md
The Zurich Bicycle dataset has been derived from the open-source Zürich Bicycle dataset developed for the original DroNet algorithm by the RPG from the University of Zürich (UZH).
Part of it is redistributed here with modified resolution, and in gray-scale to match the configuration of our ultra-low-power camera.
This release includes the same test set used in the original DroNet.
It is composed of 8564 .pgm images with resolution 324x244, each tagged with a 0/1 collision label.
Each sub-folder contains a variable number of gray-scale .pgm images, ordered by number (e.g., frame_1.pgm, frame_2.pgm, etc.) and one labels.txt file with all the ground-truth labels in the same order.